Thinkers50 Curated LinkedIn Live with Matt Beane | Learning to Work with Intelligent Machines

Matt is a faculty member in the technology management program at the University of California at Santa Barbara and is a digital fellow at Stanford and MIT. His work focuses on how intelligent machines and human beings can work together productively. Matt’s TED Talk on “How do we learn to work with intelligent machines” is worth watching and has had over 1.8 million views.

Transcript:

Stuart Crainer:
Hello, welcome to the Thinkers50 Radar 2021 series, brought to you with LinkedIn Live. I’m Stuart Crainer,

Des Dearlove:
And I’m Des Dearlove. And we are the founders of Thinkers50, the world’s most reliable resource for identifying, ranking and sharing the leading management ideas of our age, ideas that can make a real difference in the world.

Stuart Crainer:
Our belief in the power of ideas has been the foundation of our work since we launched the first ever global ranking of management thinkers in 2001. We’ve published a new Thinkers50 ranking every two years since, and it remains the premier ranking of its kind.

Des Dearlove:
So we are thrilled that 2021, a year in which fresh thinking and human ingenuity and more important than ever, is also a Thinkers50 year.

Stuart Crainer:
Nominations are now open for both the ranking of management thinkers and the distinguished achievement awards, which the Financial Time is very accurately and helpfully calls the Oscars of management thinking.

Des Dearlove:
Short lists for the Thinkers50 Awards will be announced during the summer, and then the year’s finale on the 15th and 16th of November will bring all the excitement of a new ranking and the naming of our Thinkers50 2021 award recipients.

Stuart Crainer:
Now in this series of 30 minute webinars, we want to showcase some of the freshest and most interesting ideas and to bring you the new voices of management thinking.

Des Dearlove:
We want to really inspire you to seize this moment to create a better future for you and for your organization.

Stuart Crainer:
So our guest today is Matt Beane. Matt is a faculty member in the Technology Management Program at the University of California at Santa Barbara, and is a Digital Fellow at Stanford and MIT. His work focuses on how intelligent machines and human beings can work to productively. His TED talk, How do we learn to work with intelligent machines is certainly worth watching and has had over 1.8 million views.

Des Dearlove:
So we have 30 minutes. If you’re watching this, please share where you’re joining us from today and please post your questions as we go along. Now, I’ll hand over to you, Matt. The virtual stage is yours.

Matt Beane:
Well gentleman, I’m very grateful for the opportunity. It was a wonderful surprise and an honor to be included in this really fascinating group of people. So I’ll do my best to redeem that now. I’ll take just a moment to talk a bit about me, but mostly then I want to move on to talk about what I see from where I sit and given the research I’ve done to be a very serious issue that we’re facing as a species right now, having to do with learning. Then I’m going to talk a little bit about how some folks out in the wild, despite pressures to the contrary are doing something about this, from whatever chair they’re sitting in, that I’ve found in my research. And then round it out with some thoughts about what each of us can do, no matter what role we sit in about this issue, if we care about it.

So just a bit about me. I have become known in academia to the extent that I am known as the robots and work guy. So every bit of research that I do is a generally field research. So living with, living like, sort of anthropologists style gathering firsthand data on what happens when we into reduce intelligent technologies like robots and AI, but particularly robots, into work that previously didn’t involve those technologies. What changes, what stays the same? What do we have to watch out for, and so on. I’ve done research projects, these tend to last say between a year and two years of going there, in whatever context and gathering data in robotic telepresence, materials transport sort of automated guided vehicles inside of buildings, robotic surgery was the featured tech in the TED talk, for example. So variety of kinds of technology. And one of the things that I try to do in every single study I’ve discovered over time is to find success in conditions where we would expect failure.

Basically, we have say, 90 years of research, very good quality research on what happens when we introduce new technologies into work. And the universal finding across those studies is things get worse, worse before better. And generally that is generally true, but it is also logically impossible that it’s all failure up and until a moment 10 years later, where we see success. There are people who, and organizations, who find success early in fits, in starts and imperfectly, and that’s I see as my job, as a researcher to be there at the beginning, on the ground and find the ways that not accidentally, but systematically people and organizations are succeeding with these technologies in ways that the world won’t see until later.

And why? So you could, there’s multiple sort of ways to answer that question. If you were to find me as a kid, I would have my nose in some science fiction book about robots that I was just sort of obsessed as a kid. But the other way to say this is that I was in the midst of a successful consulting career in the late 2000s, and started to read books, predicting massive change in the world of technology, all the books about exponential technology that many of us have digested. And so I applied and went to graduate school. And when I started in say, 2010, that was when the first iPad was launched. So that new hot tech at the time, and when I finished in 2017, DARPA, the defense agency for experimental projects in the US had just run a contest in which robots were competing to get in the vehicle, drive it down a road, arrive at a building, enter the building, cut a hole in a wall, turn off a valve, climb some stairs, walk across some rubble and get out the exit. This was inspired by Fukushima.

And so, the rate of progress in technology has not lived up to some of those exponential accounts, but it is quite stunning. So I think we have to be paying careful attention. All right, onto the serious issue. So everybody that I talk to immediately wants to ask and know about whether robots destroy jobs? Whether the robots are coming for our jobs? And the short answer is, they are. The available empirical research shows, and there’s limited study now, but shows, for example, in the Detroit area in recent times, when robots were implemented in organizations, for every robot implemented, 3.3 jobs were lost.

And that for those who remained, their wages went down by four tenths of one, 1%. So this sounds really serious. And there’s a couple other studies that talk along these lines. And so, at the end of the day, though, I would say we pay much more attention to that flame and get concerned about that when really we should be paying attention to an issue that is going to affect say, a billion of us by 2030. Which is not eliminating those jobs, all new technologies, eliminate jobs, and those are serious issues and we should pay attention to them. But the real issue is that these technologies, robots, AI, and so on, all of our new technologies and intelligent technologies are going to change our jobs.

And the best estimates there from say, McKinsey and other firms like that, are that between half a billion and a billion of us are going to have some aspect of our job reconfigured as a result of these technologies, them getting involved in our work. China just announced figures that imply they need you retrain 300 million people within 10 years. McKinsey’s estimate is by 2030, so that’s nine years away. Everyone in that group, let’s say a billion people, half a billion people is going to have to learn. So this to me is a much more serious issue. We need to attend to what we need to do to help us adapt to new ways of working with these technologies. And I’m going to put it to you now that school and formal training are critical, but really not the main story here, they are not the answer to our problem.

And the main answer to our problem is figuring out how to reconstitute the way that we learn while we’re doing our work, not separate from our work, not via online learning or a formal classroom or certification, but as we do our work, how we learn to do it. Right now, the news from all the research I’ve done, and then I’ve also checked across at least 25 going on 26 additional studies now done by other researchers, the clear pattern there is that we are putting all of our new technologies to use, all of these new intelligent technologies, whether they be robotics or AI, and so on, in ways that get great productivity gains when they’re successful, but at the expense of this kind of learning while doing or learning on the job.

It Has a lot of different names. In surgery, the place that I started this journey, it’s an on as see one, do one, teach one. You learn by watching a procedure for a bit, then getting involved in it with an expert present, and then they sort of coach and mentor you to work towards the more difficult parts of the work. And I challenge you to think about the most valuable skill that you have as part of the work that you do, the thing that you do that adds the most value in this world. And think about what portion of that capability came from books, even YouTube, or formal classroom training versus learning in this vicarious method, alongside an expert in some way, whether it’s editing an article, or doing surgery, if you happen to be a surgeon.

And so, you may have sort of caught the version of why we’re in trouble on that front, on the TED talk. I’m going to give you a different version now, this is from a recent study that I’ve been doing in the last couple years in warehousing. In warehousing, at least in the US, in e-commerce, the job of an organization is to manage the cost of that operation per unit shipped, basically. And the way that you do that, the primary way that you do that is to increase throughput, the number of things that go through the warehouse and reduce quality defects, right? So you want max throughput and reduced quality issues for little cost. The main way organizations accomplish this is reducing what they call skilled touches inside the building. So the fewer times your process requires a human to touch that good, that’s good.

And every time the human has to touch that good, that that touch requires the least amount of skill required, that you can get away with. Because the lower the skill requirement, the less likely somebody’s going to make a mistake. Warehouses operate on razor thin margins, so say 3%, grocery store level margins, 4%. Practically what this does, and the way that you accomplish this is to automate as much as possible. Everything you can, using advanced technology, all the way down to gravity fed conveyors, literally no computation on board. And the practical effect for someone working in one of these facilities is that they have the skill demand, their jobs demand less of them over time. So their opportunity to learn well doing, let alone learn from an expert near them, goes down over time. And if they try really hard, they’re just getting better at a skill that is easier and easier and easier to do over time. In other words, less valuable to them, their career, and to other organizations over time.

This is the net effect I’m now seeing across dozens of studies. We are taking very fine grained advantage of these new technologies in a way that sacrifices human capability advancement in the midst of the work. And so it doesn’t have to be this way, but if we leave that alone, that is the main way that we build valuable skill, and that is building valuable skill is precisely what we need to do to adapt to this new world that we’re creating with these technologies. So the sort of meta superpower that allows humanity to adapt to these new technologies that we’re deploying is the very thing that is getting weakened by the way that we’re deploying those technologies. And just in case you need a little convincing there, basically this on the job learning modality, this sort of way that we build most of our skill. If you sort of had a moment of thought, “Oh, I’m not so sure I buy that, Matt.”
I encourage you to take a look at how much we’ve spent on formal training. So in 2019, the year before the pandemic, the world spent something and close to half a trillion dollars on formal training. This is everything ranging from certifications, all the way through formal education and almost nothing, none of that went towards investigating, amplifying, extending, or enriching this vicarious mode of learning alongside experts as we work. And yet, the best available research shows that only one in five workers who are asked about the valuable skill, any valuable skill that they acquired, one in five said that they received and amplified that capability through those formal methods. So we take it for granted this on the job learning thing. And it’s, I think probably the, if there’s a message in the research I’ve done thus far, it’s that because we take it for granted, we don’t see that we are weakening it by the way that we’re handling these technologies. Now, onto the sort of rare deviance out there in the wild who are learning anyway in spite of these pressures.

I mentioned these in the TED talk, also. I called this way of learning, shadow learning because basically the people who are building skill anyway, in the conditions like the warehouse conditions that I mentioned, or in surgery, are finding new ways to build capability that run counter to what everyone locally would say is appropriate. So in the case of surgeons, those surgeons, they’re rare who build very strong, robotic capability are doing things like spending hundreds of hours with a YouTube video and, forgive the pun there, dissecting them very carefully. Or they find ways to operate with limited or no supervision by an expert in various contexts. Now they’re not doing this always deliberately, but they find themselves in circumstances where they can struggle at the edge of their capability so that they can build capability. And again, I’ve found this well beyond surgery, in many, many different contexts now, it’s a very consistent pattern.

When old ways of working are broken, and this mode of on the job learning is broken, a few of us are finding systematic, independently, finding systematic ways for which to build capability. And so those folks, I think, the best analogy I can think of is, they are the people who are walking across the lawn that a good architect watches to see where to build the path, right? So they’re not to be copied, we don’t want to follow their practices because they’re doing things that in some cases are against the law, but in other cases are just really quite untoward. And so what we can do though, is look at them and to see what are they taking all of this risk for, to build capability? Why would you operate without a senior surgeon present while you’re trying to preserve your ability to struggle in the work, which is critical for learning, not working past your capability, but close to it.
So what can we do to preserve struggle in the work, for example? My Harvard Business Review article lists a number of things for that these shadow learners are trying to preserve or amplify as they restitute their learning on the job for this new world that we’re building. So we can look at these folks and make changes, insist on changes from whatever chair that we are in. So if you are a technologist, building technology, if you are a manager, buying it for your organization, or if you’re just a user, if you’re a worker in an organization, all of us can be asking the same question. Is there a way to handle this technology that yes, gets us our productivity and results and builds our capability as we use it? This can’t always be done, but in general, I have found, we’re not really trying for this, it’s an invisible goal.

Because learning on the job is taken for granted, we don’t realize that by doing work, we’re building capability. Generally, that’s how humans have worked for millennia. And so we need a sort of reinvigorated conversation in a set of tactics around reconstituting a digital apprenticeship for the 21st century. We can’t go back to where we were, but I would say beyond just asking, “Okay, is there a way to take this technology, implement it, get our results and build capability,? As we do it. The next level is say, “How could we use these very technologies to build a new infrastructure for learning vicariously from each other as we work for the 21st century and beyond.” And I can give you a bit of yesterday’s news on this, literally yesterday’s news on my front, something that I’ve been trying to do here, which is in 2019, I published this Harvard Business Review article that sort of was a call to action on this front.

And I wrote a little love note at the bottom of that article to a guy named Juho Kim, who is also a… He and I both got our PhDs from MIT. I didn’t know him at the time, but he was doing some very interesting work around allowing a crowd of people to annotate video as a way of learning. And this looked a lot like what the senior residents were doing, who succeeded at robotic surgery, they just didn’t have the infrastructure for it. And just yesterday, a note was published in the American Urological Association’s monthly newsletter about the alpha prototype that Juho and I have been working on for the last year during the pandemic, which is now in testing to help sort of facilitate this kind of vicarious learning and joint annotation of recorded skillful work for everyone to engage in, to allow for some new infrastructure across these divides that we are sort of inadvertently creating in the work, so that we can all learn from each other, whatever our expertise level is, as long as there’s a repository of skilled work.

Now, that’s just one example of taking technology to address the problems and building a new kind of technology to address the problems that are created by the way that we’re handling the old, or it’s not really old, last year’s technology. There are a universe of ways in which we could build new tools with the very tools that we’re all complaining about to address the challenges made by the previous. And so I think we need, especially given our need for adaptation now is increasing, the world is changing faster, we’re introducing… And I don’t think we’ve really seen anything yet in terms of what AI can do and how it will transform the world, I think we are just beginning to see that.

And so if we’re having all of these challenges now, in terms of weakening the main way that we adapt as we do to new technologies, new ways of working, it needs attention now, at least that’s my point of view, that’s what I’m going to be devoting the next leg of my career journey to. I invite you to the journey, and gentlemen, I’m grateful for the time. That’s basically the remarks I had prepared.

Stuart Crainer:
Brilliant, Matt. Now, usually when people are speaking, I write down questions based on what they said. With you, I just found I was writing down what you actually said, because it needs kind of unpacking in my head. So, who is… So it’s [inaudible] it’s a race to reinvent learning, is what we’re engaged in. And so the obvious question from that is who is leading the race. I mean, by the sound of what you say, if China is dedicating resources to retraining 300 million people.

Matt Beane:
Well-

Stuart Crainer:
[inaudible] it’s clear.

Matt Beane:
So no one. I think if you buy my argument up till now, no one is the correct answer. I’s a bit of hubris I know, but we are all focused on and investing quite heavily in, we should be investing perhaps a lot more in formal learning channels, non credentialed kind of things like Khan Academy and Coursera and sort of, and alternative ways of getting employment through training. There’s a variety of organizations that have started that as well, but all of these are focused on reinvigorating and building up the formal ways that we learn, separate from work, a digital class, an in-person class, asynchronous, synchronous, these are critical and yet not the bottom of the pyramid, so to speak.

I think we’ve sort of got our attention flipped. And so I don’t, I mean, I scour the web every day for news about learning on the job, apprenticeship programs, mentorship programs, and the number of web hits there compared to anything to do with formal curricula and training, it’s just tiny. There’s some inspiring examples there, and then again, those are just traditional programs. And especially in the world we live in now, a traditional apprenticeship program needs a little bit of digital infrastructure, we need to be bringing these tools to bear on that problem in the same way that we’re trying to work in comparable ways that we are trying to do with formal skills development.

Des Dearlove:
Okay. Just go back a step. What I’m curious to know is, is this a new problem? Because we had de-skilling with previous technologies, division of labor, the assembly line, Ford. We stripped away people’s skills and at least that’s what we were told. Is this something radically new that’s happening now?

Matt Beane:
No, no. Yeah. I think that what is different is the intensity of the problem. So you can go back to the Luddites, you can even go back, there are lovely stories of sort of comparable revolts amongst pottery makers in Ancient Greece at the advent of the spinning pottering wheel. So before that, it was all made by hand, it was very labor intensive, you were laying coils on top of pottery. And then someone invented this pottering wheel and there was this massive conflict, there were concerns about de-skilling, we even have written records. And so anyway, no, that is as old as civilization is problem, it’s just that the velocity, and intensity, and the breadth of the change back then was sufficiently slow that three generations later at the pottering wheel had supplanted coil techniques.

And that’s plenty of time for people locally to adapt. What I’m seeing now in the data, and the studies that I’m doing is that the velocity intensity and the sort of diversity of fronts on which this is happening is greatly increased, and that is partially because, and I’m sort stuck in surgery, so forgive the pun, we have a much finer scalpel to dissect work with. The new tools that we have, allow us to separate an expert from an apprentice in the name of efficiency, much more carefully, so that we can amplify and extend that expert’s capability. Why wouldn’t we want to do that? Well, the hidden cost of that is that that trainee is cut out of that interaction. So we’re able to do it much more efficiently now, is my answer. So same phenomenon, but yeah, I’m repeating myself now, but that’s the basic answer. I think it deserves more of our attention, it’s also global now.

Stuart Crainer:
There’s a question from Hal Gregson of MIT. “What are the one or two new capabilities that C-suite leaders specifically should acquire to tackle the challenges that you have outlined?” Matt.

Matt Beane:
Well, so, hi, Hal. I welcome that question and I’ll answer it with a question, because I know you’re a fan of questions, which is how can we get our productivity and capability too? So forget a thing to do, how about a way to think? A way to think is, we have been sacrificing inadvertently, building the capability in the organization, challenging it, stressing it, and keeping humans sort of where they are best, which is as learners, as adapters. We’ve been sacrificing that at the expense of trying to go for productivity, that’s an optional trajectory and I would say not ideal. So if you’re a savvy CEO, that question should be burning in your mind. How can we handle these technologies in ways that enhance the capability of our people and get us results? You locally, as a CEO are going to know the answer to that question, or you’re going to have means of getting an answer and it will be yours and your organization’s alone, it’ll be unique to you. But not asking that question is a dangerous thing.

Des Dearlove:
A couple of people commenting on your phrase, “Find success in failure,” or “Try to find success in conditions where we would expect failure,” which is a fascinating thing. I seem to remember though, I’ve seen you talking about bomb disposal. There are examples where people are actually using digital technologies in a different way. I know that’s what you’re advocating. But can you talk about that a little bit, because it seems like a particularly good example?

Matt Beane:
Yeah. So I’ll go bomb disposal, and then perhaps to something less intense, which is working in warehouses, that’s been a recent example too. So bomb disposal is the only example that I could find looking across every study I went through. Dozens and dozens of recent studies to find an example where, because we deployed these advanced technologies, learning on the job got better. And the twist there was that the pre robot condition for bomb disposal was, if I’m a senior bomb tech, and you’re a junior bomb tech, and we’re in a truck, bombproof truck, the way we handle this situation is that I get out of the truck in a suit and walk over to the bomb and poke it with a stick, and we are 300 meters apart. You’re lucky if you have a high quality video feed of what I’m doing. In general, no.

And so you only as a junior person learn about what I’m doing through my talking about what I’m about to go do. And then if I’m lucky enough to come back, telling you what I did. Right? And so, what happened with the robot then is that we could send a robot and you as a junior tech would control this robot, I would direct you to control it, drive it up to the bomb, and I would talk you through my mental process as we’re doing that, you would use the robot to diffuse it, I would direct you to do that, if we are lucky enough to diffuse it, and then if we couldn’t, then I would have to go in, but at least then you get much more of the diagnostic window, and I’m sort of talking you through walking you through the work as I’m doing it. So it’s better now with robots.

There are some examples in warehouses for frontline workers, these folks that you would expect to be most de-skilled and most trapped in losing capability over time as a result of the introduction of this automation and we’re just about to submit a paper off to a journal about this after two years of work in that department. And I can say that the core insight that we’ve gotten there is that those people who do get ahead are doing so because they are finding ways to note trouble, problems in this ostensibly, perfectly automated work and to fill the gap with creative effort, to make the system function in ways that is pretty invisible to management, pretty invisible. And in general, it doesn’t get you ahead, but these are the folks building the capabilities required to enable advanced automation for the 21st century. They’re mostly… And so they’re not visible yet, they’re not getting promoted, they’re not getting more paid, but they are building skill.

Stuart Crainer:
Are you optimistic, Matt?

Matt Beane:
In general? Sure. Yeah. In general, yes. Right? The bigger the threat, the more humanity unites to address it. And I think that applies not just to the issues that I’m focused on, I think that’s what we can look at and see over the span of human history. And if we are generally under threat in terms of our ability to adapt, I think we’re going to do something about it. The alternative is not palatable, and we’re incredibly inventive with these technologies. And the young kids I see now trying to build new technologies, have such purpose driven goals, they really want to make the world a better place, they’re not after money only. So yes, absolutely. There’s much work to do, and of course, failure is always an option, but yes, definitely optimistic.

Des Dearlove:
Okay. I’m afraid we’re running out of time. This is one of those topics that is just, the more you get into it, the more fascinating it becomes. Matt, thank you. We will resume this conversation another time I think. But we are out of time. So please join us again same time next week.

Matt Beane:
Thank you very much, gentlemen.

Stuart Crainer:
Thanks very much, Matt.